60105 - Grid Plate Flame Stabilizer for High Intensity Gas Turbine Combustion: The Influence of the Method of Fuel Injection on Mixing, Flame Development and NOx Emissions 

Grid plate flame stabilisers for low NO_x emissions have renewed interest in recent years due to work on FLOX combustion and their use as the basis of low NO_x hydrogen gas turbine combustion by several companies.  For non-premixed grid plate combustion the differences in the literature in the use of grid plate flame stabilisers is in how the grid plate air flow is fuelled. FLOX burners inject the fuel on the air hole centreline, hydrogen grid plate burners inject the fuel radially inward into the air flow. The authors have investigated radial inflow fuel injection using 8 fuel holes for each air hole (Grid Mix 1, GM1) and fuel injection through an annulus around the air hole exit (GM2). The radial fuel injection was shown to achieve some of the lowest NO_x in the literature. The GE hydrogen combustor and the micro mix design developed by U. Aachen and Krebs, use this technology. This work uses ANSYS FLUENT CFD predictions of GM1 with comparison with axial gas composition traverses inside the combustor. Also the FLOX mode of fuel injection (GM3) on the air jet centreline was examined using CFD to determine if this was the best fuel injection location. For all three fuel injection modes the recirculated mass flow of burnt gases was computed. Previous CFD work of the authors for non-reacting fuel and air mixing showed very quick mixing for GM1 and the slowest one for GM3. So the intention of this work is to corroborate these predictions that should lead to GM1 having the lowest NO_x emissions . A 76mm diameter combustor, with no film cooling, using a four hole grid plate flame stabiliser with 19.62 mm hole diameter was used for this study, with a plate thickness of 9.52mm, and a design pressure loss of 2% at the reference Mach number of 0.047, which is the flow condition for all the combustion air flowing through the flame stabiliser. The air inlet temperature was 400K, as in the experiments. The thermal input was 131 kW at Ø=0.6and the heat release was 28 MW/m², which was a very highly loaded combustor for all the combustor air flow through the flame stabiliser. The predictions were for an equivalence ratio in the range 0.5-0.6, as in the propane experimental work, and the predictions were also carried out for hydrogen  for both fuels at 288K. The simulations were carried out by sing RANS simulation, k-epsilon turbulence model, and the non-premixed combustion model using a PDF approach. The NOx was calculated as a post process computation, using a NOx model included in the software FLUENT. The results confirmed the results of the mixing predictions that the lowest NO_x occurred for the most rapid mixing system the GM1 design. Reasonable agreement with the axial gas analysis traverse results was shown, but the predictions had a slower flame development than in the experiments. The averaged exit emissions were in reasonable agreement with the exit mean gas sample probe. The hydrogen predictions for the same mean temperature conditions were for a faster flame development and a small increase in NO_x due to the greater residence time in the combustion gases.